Author: parsec
Subject: SM951 as OS boot device information
Posted: 31 Aug 2016 at 12:21pm
A few of those links? The only one I recall is HWBOT:
http://hwbot.org/
How is X99 outdated? You can use PCIe NVMe SSD directly on the PCIe 3.0 lanes from the CPU, rather than the DMI3 lanes from the chipset. Is DMI3 faster? No.
Sure RAM caching or a RAM drive like ASRock's XFast RAM, is faster than any SSD. DDR3 and DDR4 memory is faster than NAND storage, nothing new there. They've been around for years, so why aren't they more commonly used?
RAM caching and RAM drives are not permanent storage. Remove power from the PC, and their contents is gone. Or if configured right, written to a SSD or HDD, at the speed of those devices, when the PC is shutdown. But if a power loss occurs, any data in a RAM cache or RAM drive is lost. RAPID has the same limitation.
My complaint about RAPID is some people seem to believe it is the SSD providing those speeds, when it absolutely is not. If you were bench marking SSDs as a hobby, would you want your efforts compared to RAM drives or a drive using a RAM cache?
The only data stored in RAPID's RAM cache is new data "written" to the RAPID enabled SSD. That data is written to the RAM cache, and then written to the SSD. If you open an older file on a RAPID system, it won't be read at the speed shown in a RAPID enabled SSD benchmark. The benchmark test data is in the RAM cache, not the SSD. The joke is, after the benchmark test is done, the test data would be written to the SSD, if the benchmark program did not delete it, as it normally does. Is any writing to the SSD done at the speed shown by a benchmark run on a RAPID enabled SSD? No, it is written at the speed the SSD itself is capable of. If anyone thinks RAPID makes their Samsung SSD faster, they are wrong.
When a RAM drive or RAM caching is used, when the PC is shutdown or restated, the data in the RAM drive or RAM cache must be written to a standard SSD or HDD first. Otherwise, it is lost.
Why doesn't the file system of an OS do that automatically now? Cache data to RAM, and then write it to permanent storage behind the scenes, as RAPID does. Sounds simple, right? Write caching exists in Windows now, and in the IRST AHCI and RAID driver. But the size of the cache is small, for many reasons.
Subject: SM951 as OS boot device information
Posted: 31 Aug 2016 at 12:21pm
 ReRoR wrote:Hey parsec, Although some sites include RAPID mode in their results, and it may be considered cheating the benchmarks, if they specified the hardware platform, and that RAM caching is enabled, it may have some actual purpose for users, and could also be a real way to fetch some awesome performance gains for gaming and applications, with much lower outlay on technology. Would you mind kindly sharing a few of those links for me to have a look at closer? Thanks! I am particularly looking to exploit this using an outdated X99 chipset for a build, and it's quad channel memory, as you can see why;
There are also 3rd party applications that perform similarly to Samsung's RAPID, that may be introducing the ability to define what specific information is cached, as opposed to a straight bit level cache everything, which would further add merit and purpose to the use of RAM caching. |
A few of those links? The only one I recall is HWBOT:
http://hwbot.org/
How is X99 outdated? You can use PCIe NVMe SSD directly on the PCIe 3.0 lanes from the CPU, rather than the DMI3 lanes from the chipset. Is DMI3 faster? No.
Sure RAM caching or a RAM drive like ASRock's XFast RAM, is faster than any SSD. DDR3 and DDR4 memory is faster than NAND storage, nothing new there. They've been around for years, so why aren't they more commonly used?
RAM caching and RAM drives are not permanent storage. Remove power from the PC, and their contents is gone. Or if configured right, written to a SSD or HDD, at the speed of those devices, when the PC is shutdown. But if a power loss occurs, any data in a RAM cache or RAM drive is lost. RAPID has the same limitation.
My complaint about RAPID is some people seem to believe it is the SSD providing those speeds, when it absolutely is not. If you were bench marking SSDs as a hobby, would you want your efforts compared to RAM drives or a drive using a RAM cache?
The only data stored in RAPID's RAM cache is new data "written" to the RAPID enabled SSD. That data is written to the RAM cache, and then written to the SSD. If you open an older file on a RAPID system, it won't be read at the speed shown in a RAPID enabled SSD benchmark. The benchmark test data is in the RAM cache, not the SSD. The joke is, after the benchmark test is done, the test data would be written to the SSD, if the benchmark program did not delete it, as it normally does. Is any writing to the SSD done at the speed shown by a benchmark run on a RAPID enabled SSD? No, it is written at the speed the SSD itself is capable of. If anyone thinks RAPID makes their Samsung SSD faster, they are wrong.
When a RAM drive or RAM caching is used, when the PC is shutdown or restated, the data in the RAM drive or RAM cache must be written to a standard SSD or HDD first. Otherwise, it is lost.
Why doesn't the file system of an OS do that automatically now? Cache data to RAM, and then write it to permanent storage behind the scenes, as RAPID does. Sounds simple, right? Write caching exists in Windows now, and in the IRST AHCI and RAID driver. But the size of the cache is small, for many reasons.